Will There Be A Winning Autonomous Robot in 2005?

An anonymous reader submits "This summer is heating up the DARPA Grand Challenge as multiple top notch schools begin to announce their entry into the competition. The newest organization to announce its entry was the Florida Institute of Technology. Their project is known as Oasis - Autonomous Racing, and they have a team of over 45 students, professors, and advisors that are currently hard at work designing their vehicle and raising funds to pay for it. The DARPA Grand Challenge is a race between vehicles that should be designed to travel up to 300 miles in less than 10 hours through the desert or other harsh medium without any human interaction. The 2005 competition has a $2 million grand prize as authorized by congress. With all of the new entrants does anyone think that the competition will be won the second time around?"

Comments

Doubtful

[7Ghent]

Well, considering that the best performer for this year didn't even make 15 miles, I'm hopeful that someone will actually complete the course, but not in under 10 hours.

Re:Doubtful

[wakejagr]

how about an award for getting past the first turn? that first left turn was too much for quite a few of the contestants in the first challenge.

Re:Doubtful

[ciroknight]

I don't doubt it for a minute. Being a Florida Tech freshman in the fall, I want to try to get involved with this project. I've worked on stereo optics systems before using two webcams and I can tell you that this kind of system holds great promise at winning the race. In combination with great laser range-finding and possibly optical range finding (something like us humans do), even acustical systems, a machine has just as good of a chance to pilot a car as it can an aircraft or anything else.

Another thing: use the 2d images to build a 3d map on the fly, approximenting object sizes by finding the edges of the object in the pictures, and you should be able to navigate around and over them quite easily. The car also plays a key roll; it needs to be adapted into a dune buggy of sorts; huge soft tires and great suspension.
Scary that we're working on this for the government though..

Re:Doubtful

[Jack Porter]

Not sure where you went to school, but IIRC 300 miles in 10 hours is 30 mph.

Re:Doubtful

[rmohr02]

Ohio State had a team that entered this year, and they worked on their entry right next to the workspace of a project team I'm a part of. I heard, albeit secondhand, why most entries didn't work, and the only real reason is that the teams were unfamiliar with the course (e.g., Columbus doesn't exactly have a large desert to test the entry in).

A couple nights before the competition was to take place, it rained on the "course", as it is. Thus, there were many relatively large bushes in the desert when the competition started. This was not something most teams had planned for--however, they did plan for large rocks. Thus, a 9' tall truck would drive up to a (now relatively small) bush, detect it, determine it was a rock, and then try to plot a course around it rather than simply driving through it, which would have worked fine. With the number of bushes that had sprouted up, it was only a matter of time before a truck's computer got swamped trying to avoid all of the "rocks".

I look forward to hearing about next year's competition, for which I'm sure teams will think to find a way of differentiating a bush and a rock.

Re:Doubtful

[shigelojoe]

He might work for NASA. ; )

Re:Doubtful

[Nugget]

At first I thought that this post represented a new slashdot low (Math is hard, let's go shopping!) but then I noticed that the damn thing has been upmodded twice.

JPriest may just be having a caffeine-free day, but who are the two jagfucks who thought this was interesting and insightful?

Re:Doubtful

[renec]

Is that the new math?

You know how to convert to metric, but when you divide 300 by 10, they get 60. I'm assuming you asked google about the metric conversions.

The average speed is 30 mph, or a bit under 50 Km/h

Re:Doubtful

[homer_ca]

That's the perfect example of something that's easy for humans and hard for robots. It's one thing to detect the presence of obstacles. It's another to identify obstacles and determine their risk to vehicular progress. We know oil is black, shiny and slippery. We know rocks usually look jagged and the color of dirt, and they're hard when you run into them. We know that it's a long way down if you fall over a cliff ledge. If there's a cliff wall going up on one side of the road and down the other side of the road, we know that falling off the cliff is much more dangerous than hitting the side of the mountain, at least at low speeds. It's common sense things like this that humans just know, and it's hard to program every possible scenario into an AI.

Re:Doubtful

[Quixote]

Another thing: use the 2d images to build a 3d map on the fly, approximenting object sizes by finding the edges of the object in the pictures,

Try finding the edges in a bush or a clump of tumbleweed...

Re:Doubtful

[murmurr]

Ummm, those bushes don't spring up overnight. Not to more than a couple of inches, anyway. Nevertheless, since there was some sort of "roadway" for the entire length of the route, there was really no need to distinguish between rocks and bushes. If your vehicle was intended to steer around a 2" rock, then you had made a fatal mistake before leaving the gate.

Re:Doubtful

[ciroknight]

Trust me, I understand that too, but that's when you have to get into surface approximentations and such. If a surface looks scattered like a clump of grass, then it's booleaned from a screen of black. The resultant image should be easy enough to pattern match to images of trees, rocks, brush, or anything else in a library. Dunno how well this would work in real-life, but this is what I did with my little toy model and it worked well.

Another way might be to test the density of the object, or to use color like we humans do.. If somethings green or yellow and sparce, it's more likely tumbleweed or a bush, otherwise it's a rock.

Lastly, at certain speeds, objects of certain sizes should be tossed out. If it's the size of a small mellon *smaller than a water mellon lets say*, then just throw the object out. It's not going to effect the car going at a speed of 35-65 MPH if the car's built right.

Just a few ideas. If I do get the chance to be in the project like I'm hoping, I'll get to test a few of them. Otherwise it's probably just idle chatter. Either way, it's something to think about.

Re:Doubtful

[ciroknight]

I usually keep it as a policy not to respond to AC's; they don't have the balls to say it to my face, then it means nothing. But this time, I'm going to reply anyways.

Yes, there are cases where there will be no edges to find, but in those cases, it's usually okay to just keep trucking a-long. It's not just one technology that will win this race, it's the combination of all of them working together to solve a common problem.

As for speed, increasing the shutter rate of the camera, using multiple cameras, and faster computers deal with speed processing. The idea is to slowly ramp up speeds as it's safe, and the instant there's something in the way, begin to slow down and steer, noting the deviance against the path, and undoing it as soon as is convienient.

To end all, yes, the real world is different than the theoretical and the experimental world, but both worlds are based on the real world, and the rules are the same. Approximenting is what we usually do in the real world to make up for the inadaquacies of it. Since the world isn't perfect, neither can our machines be.

Re:Doubtful

[1iar_parad0x]

Of course, I hope "smart" vehicles don't just learn to identify obstacles merely by their risk to vehicular progress. If these robotic cars ever plan to be useful, they're going to have to learn not to run over things just because they can and it is computationally more efficient.

This does remind me of that "made for TV" Knight-Rider reunion where Michael's new car ran over a deer because it was calculated to be more efficient that slowing down. We all know that David Hasselhoff is a great actor.... Actually, I can't even write that in sarcasm. Wow!

For God's sake..

[GeneralEmergency]

...I hope nobody names their AI unit "SkyNet".

Now, where did I leave those keys to the bunker?

Short answer ...

[Manip]

No.

I don't think anyone will win this time around. The problem is that the current technology can't deal with unknown situations/objects, maybe in a controlled enviroment with selected things added and removed but in a desert there is very little chance. If someone does win it will be more down to luck than actual computing power.

Re:Short answer ...

[ejaw5]

I think if you do a bit of research, you can find microcontrollers and the sensors needed to accomplish the task. Not to simplify the Grand Challenge, the objective is to have a vehicle traverse through a desert terrain while avoiding other vehicles and obstacles. Given enough time, any good electrical enginnering student(s) can come up with some good ideas on solution with some possible hardware choices.

The Challenge to DARPA isn't the technology, but the testing phase, or lack there of. How many of the schools who participated last year had practical access to a desert of similar circumstance? (I'm in north Florida and I can't think of a place) You can put together an autonomous vehicle for the competition, and maybe test it in a large open field with some 'simulated' obstacles but won't come close to the real deal.

Another thing is 300+ miles might be pushing the limit for how far gasoline vehicles (especially the trucks and SUVs) can travel on one tank of fuel. Keeping in mind that there would most likely be frequent go and brake driving, if any vehicle were to make it across the finish line the fuel gauge would be below the slash.

Re:Short answer ...

[Waffle Iron]

The problem is that the current technology can't deal with unknown situations/objects, maybe in a controlled enviroment with selected things added and removed but in a desert there is very little chance.

What it boils down to is that there's something horribly wrong with the current approach to "AI". Nature solves problems very similar to this with a totally different approach. Take a cockroach for example. Its task is probably much harder than this "grand challenge". It must survive in the world for several weeks or months while: finding its own fuel, avoiding hostile predators, finding a suitable mate, and include a control system that supports walking in any orientation along with controlled flight through the air.

What computing horsepower drives this task? A few milligrams of wet neurons that probably consume a few microwatts.

Even if a cockroach weren't up to driving one of these vehicles through the desert, any small bird probably has enough signal processing power to handle the chore. They certainly are able to handle flying through a thicket of tree branches, a pretty tough challenge in itself. How much does a house finch brain and vision system weigh? Maybe 1 gram?

Back in the 80s I majored in AI briefly, and I quickly came to the conclusion that the incredible pattern matching abilities of living organisms can't be effectively modeled by piping numbers through a single accumulator register. The highly interconnected architecture of a brain is totally different. (Many of my professors seemed to think that they had some deep secret insight to "intelligence" because they were hacking in Lisp. What was really happening was that they were caught up in their own cleverness in using recursion and macros to create layers of abstraction. But that's just tricky discreet math, not self-awareness.)

Now that computers are 1000X faster, my assessment is still valid. In fact, computers probably aren't even nearly 1000X faster at the algorithms that living organisms use to deal with the real world, because all of the computer speed tricks rely on locality of reference (caches). A brain, OTOH, is a fully associative processor that can compare an large chunk of input with a good amount of its entire memory in a single atomic operation. Its power comes from not having locality of reference.

IMHO, attempts at these kinds of projects are always going to result in clumsy, kludgy, stupid machines until some totally new approaches are developed for processing and information retrieval.

Re:Short answer ...

[Wog]

And these design differences, my friend, are what seperate the mind of man from the mind of God.

Burn, karma, burn!

Re:Short answer ...

[pauldy]

Man, why is this modded as troll? Are there slashdoters out there that are so intolerant of the use of the word god.

It's not about winning...

[Wiser87]

The whole point of the race is to see how well non-government groups solve these problems and to gain new insight on how to use technology.

Just getting something that works makes them winners.

Going out on a limb...

[DocJohn]

I know we're going to hear mostly naysayers here, saying "Well, gee, they couldn't even make it 15 miles this year, what's the chance of anyone actually winning in a year's time!?"

I think there's a good possibility that someone can win it. Think about it. This past year, none of the teams had any first-hand, direct experience with this course or the challenge. So now every team has all of the experience and data from this year's challenge, and could not only see what went wrong with their team's entry, but the problems faced by every other team (motorcycle entry notwithstanding).

I think the computing power is there. If the teams learned anything from this year, it should be that GPS isn't sufficient in and of itself. You need to far more creative. Every system should have 2 or 3 redundant subsystems.

I think it can be done, and I think there are enough creative people working on the problem that it wouldn't surprise me to see a winner next year.

Short answer...

[sinner0423]

No. Last time a good 30%, I believe, didn't even make it out the gate. I seriously doubt any of them will "win". Well, I'm sure they're all winners, like in the special olympics, but i don't think they will FINISH the course.

With the closed nature of the competition, no.

[dexterpexter]

If the prior entrants are any indication, than no. Those entrants shows just how unprepared they were. As a engineering student on a team that has built/is building an autonomous robot (not associated with DARPA), my evaluation of the vehicle designs left me terribly disappointed. In fact, part of me things my own team could have thrown our present navigation hardware/software onto an ATV and been more competitive than the other DARPA entrants. In fact, had DARPA not been so selective in their choosing of robots to enter the competition (which, in my opinion went against the spirit of an open competition), we might have done just that.

A few responders have said that the technology just isn't there for autonomous navigation. I disagree. It just needs to be refined. Robots for the IGVC can navigate unknown environments respectably, and these are unfunded, poorly staffed projects ran by undergraduate students.

I believe that the next competition's entrants will make it much further than this years, but looking at the stock, similar designs that DARPA let through, looking at bells and whistles rather than creativity, my hopes are not high for having a winner. They need to re-evaluate the meaning behind an "open" competition of ingenuity and consider that the most expensive, technologically-advanced robot is not always the answer.

Look at the first year IGVC. Colleges spent thousands of dollars on big, relatively the same robots and the University of Tulsa came in with a PC bungeed to a child's car and beat them all. I don't pretend that the IGVC robots are competitive against the Grand Challenge ones, but the point is still the same: make it an open competition, and perhaps we might see some *real* ingenuity and then, in the future, a winner.

Money d.n.e. ingenuity

That said, I tip my hat to the previous entrants. How neat is this competition!? (even with its limitations)

Highly doubtful

[OblongPlatypus]

According to the May issue of Wired, the best team got through only 7.4 of those 100 miles before breaking down. There are some funny quotes in the Wired article, showing just how miserably far away we are from true autonomy:

What went wrong: "Lost GPS signal. Forgot there was a mountain between it and next checkpoint. Tried to drive through mountain."

Lesson learned: "Go around mountains, not through them."

What went wrong: "Interpreted small bushes as enormous rocks and repeatedly backed away from them."

Lesson learned: "Get new sensors that can distinguish between bush and rock."

This all sounds pretty pathetic, but having just completed a master-level course in artificial intelligence, I suddenly understand just how difficult some of these issues are to solve. Let's face it: We won't see anything even approaching true autonomy in anything but tightly controlled environments for years to come.

I conclude with the best quote; not really AI-related, but still simply hilarious:

What went wrong: "On-off switch located on side of vehicle. Bumped into a wall on way out of start area. Turned self off."

Lesson learned: "Put the on-off switch somewhere else."

How hard would it be...

[JawnV6]

to hide a Little Person inside one of these things? Baron Kempelen got away with such a scheme for quite a while... The Turk

It might happen...

[sgtsanity]

BTW, for all those interested, Wired ran a list of what went wrong for each team. It reads very comically, but a lot of these things are very "DUH!" after you've gone through the first time. I forsee a lot better results, as teams will have that much more practice. Hopefully some will come up with some more general solutions, rather than brute-force processing the terrain around the known area of the route.

Re:It might happen...

[Anonymous Coward]

None of the teams managed to do much brute-force terrain processing, because in general, their sensors didn't work.

The one that got the farthest just ran off of pre-computed GPS waypoints, and as the GPS accumulated drift error, it started driving to one side of the road, then in the ditch, then off road, until it hit something and stopped.

Against the rules

[dexterpexter]

You can't use a hover-craft, though. I am sure that is the first thing on *everyone's* mind. (I know it was mine)... why not just build a helicopter and make it take the most direct route? There is a reason no one did that:

The rules limit entrants to mechanical-to-ground travel. No hover crafts allowed.

However, there are other non-DARPA competition where flying autonomous bots are preferable. DARPA's competition, however, is limited to road vehicles.

Re:Against the rules

[Maserati]

If you're going to go tethered, then why not just go inflatable for the remote unit ? Other than snagging cable you won't have too many problems from having a balloon floating over your robot. It just needs to be bouyant enough to lift a small camera. I'd suggest using compress-air cartridges to inflate the balloon, you probably don't want to hassle with any sort of compressor.

Correction - number of miles

[OblongPlatypus]

I mean 142 miles, not 100. 142 is the number Wired quotes.

The /. blurb mentions 300 miles, but the Q&A on the DARPA page says "will not exceed 300 miles". Apparently the course is randomly selected and only revealed on the race day, to make sure the vehicles aren't trained for the specific race course. I'm assuming the Wired quote means that the course that was picked for this 2004 challenge was 142 miles long.

rallying

[SuperBanana]

Lesson learned: "Get new sensors that can distinguish between bush and rock." This all sounds pretty pathetic, but having just completed a master-level course in artificial intelligence, I suddenly understand just how difficult some of these issues are to solve.

Watch a rally. Rally drivers have codrivers w/notes, and prior knowledge of the course...but I believe with Baja it's mostly seat of the pants; Paris-Dakar has got to be since it's so damn long, but I could be mistaken. They average well over 60mph on a course that's got to be much worse than anything DARPA came up with. Of course, they have astronomical component failure and driver error rates (as well as the occasional wildlife incident- one rally team hit a cow at well over 60mph, it was NOT pretty- I think they also got arrested, because it was a serious crime in the host country, akin to murder, to kill a cow), and at 60mph, rocks look like bushes and bushes like rocks, until it's way too late to do anything about it. Rally teams just bolt up more plating on the important stuff, and hope for the best.

What went wrong: "On-off switch located on side of vehicle. Bumped into a wall on way out of start area. Turned self off." Lesson learned: "Put the on-off switch somewhere else."

While not defending them, it was probably an emergency disconnect switch, which you do want to be highly accessible for those times when, say, it starts driving away (or towards something) and shouldn't have. Yes, DARPA required radio safety switches, but do you really want to trust your life to just a radio disconnect?

Honestly, some teams were just stupid in their use of money and priorities- I got a huge kick out the team that had a giant plasma display TV in the passenger side of the cabin. What the -fuck- was that for, watching the Superbowl while the car drives you to the next checkpoint?

Yes.

[simetra]

His name: Al Gore.

Thanks, I'll be here all week. Be sure to tip your waitress.

The real question

[commodoresloat]

is whether the winning robot will have a "CAPS LOCK" key.

Slashdot entry

[Anonymous Coward]

Surely with all the talent on Slashdot, we could create a winning entry?

Name suggestion: The Autonomous Coward

Baja

[crisco]

Depending on the class, the Baja (and other desert races) contestants depend heavily on co-drivers, GPS and proper preparation. They run over the entire course before the race (hence the 'PreRunner' style of trucks) and rely on maps, GPS and the co-driver's experience.

Motorcyles and the trophy trucks averaged nearly 60 MPH on the last Baja 1000, other classes are slower.

I wish Rally driving were more popular over here in the US of A, so much more excitement than big ovals.

prediction: the winning entry will be D6 based

[bandy]

I predict that this will finally be won by a Cat D6 (bulldozer) based vehicle. Drive through small things. Don't get caught up in barb-wire fences. A little GPS and some vision thing for detecting deep holes and you're there.

Most of them will never work

[Animats]

With the possible exception of CMU, nobody had a system that could avoid a ditch or a pothole. Stereo vision won't do a good enough job on dirt for long range ditch/pothole detection. All the laser rangefinders except CMU's were fixed line scanners, so they couldn't possibly profile the ground ahead reliably from a bouncing vehicle.

CMU's approach is a big hammer. They took a stock line-scanning laser rangefinder and put it in a huge 3-axis gimbal, which they then actively stabilize. That should be able to profile terrain, but it's a huge mechanical kludge. If you miss a spot because you hit a bump, you have a hole in your data. At that point you can either slow down and rescan, or plow ahead blindly. They may eventually complete the course with that rig, but no way is it a commercially viable technology.

The next generation of sensor technology may be ready in time. There are at least three groups with usable sensors in the prototype stage. We're talking to two of them. But that's all I'm going to say for now.

John Nagle / Team Overbot.

(We're recruiting. See our jobs page. No pay, some risk, a fraction of the prize, we cover all expenses. Silicon Valley only. We have our own shop in an industrial park in Redwood City. If you're local, come over and see the thing.)

Re:Most of them will never work

[Animats]

Why not a combination of stereovision, range finding, and a digital horizon to enable real time mapping based off a visual system?

Stereo vision has two fundamental limitations. First, it doesn't work very well unless the scene has clean, sharp edges to match up. Second, the accuracy decreases rapidly with range, beause you're measuring a narrow triangle from angles at the base.

The algorithms for stereo vision aren't all that forgiving. There are basically two flavors. One finds and matches "features", usually corners. This works nicely for indoor scenes and badly on dirt roads. The other does a straightforward correlation between matching scan lines from two cameras, sliding them back and forth looking for the best match. This has a high false alarm rate on surfaces with high-frequency detail, like gravel roads.

Practical problems include the fact that correlation algorithms are sensitive to high-frequency noise, so any thermal noise from the camera is a major problem. Also, keeping two cameras aligned to within a pixel while jouncing along on an off-road vehicle requires a very rigid mounting with the cameras near the center of gravity along the inter-camera axis. (For an example of a good one, see the Bumblebee from Point Grey. They have the most successful stereo vision products.)

To date, the most successful outdoor stereo vision system used on a mobile robot was on the NASA Hyperion robot. They were able to achieve a range of about 7 meters on rocky terrain with hard edges. This is about a third the range that theory predicts. A DARPA Grand Challenge vehicle needs at least 20 meters of range, and if you want to go fast, 50 meters. You need 1.5 to 2x your stopping distance.

We have a stereo camera setup working on my desk here, and we've had it for over a year. We've tried that.

Stereo from motion, where you work with successive frames from a single camera, has potential. The baseline is the distance you move between frames, which can be much bigger than the distance between two cameras. But people have been trying to make that work for years without much success. If you want to work on vision, that's a good problem. Especially since you can just take data from a camcorder and crunch on it - no special hardware required for development.

Re:Most of them will never work

[rebelcool]

They're getting better, but its still years away from usefulness. This past week I checked out some new work that used temporal data streaming to fill in whats between the discrete frames (much more like how human vision works). This allowed much more detail to be considered, with lower noise errors that plague stereo vision traditionally.

However it still required structured light to work well, meaning outside the lab it wont work well.

One problem that I see with stereo vision research at the moment is that its still focused on turning cameras into fancy rangefinders. Though its nice to be able to determine the entire range of objects in a camera's vision area (as opposed to a singular laserbeam endpoint), it still tells you very little about the nature of those objects which is a great deal more useful than their range.

I don't think that a balloon will work here

[dexterpexter]

A balloon is a fine idea for slow-moving robots.

However, towing a balloon behind a robot that travels at an average 30 mph would present a problem.

For a demonstration:
Fill a balloon with helium and then try to run with it. Instead of staying afloat, it will sink.

Then there is another problem. Compressed air cartridges would only dispense AIR into the balloon.

You cannot simply fill a balloon with air and have it float. You would need something like helium.

But then, you still run into the issue of trying to manage a balloon at high speeds.
It would work if your strategy was to stop and then release the balloon, then retract it before resuming. Problem with this is that a balloon would be more subject to the wind (deserts are notorious for horrible winds), accidental tears in the bushes, and a lack of stability (what is to stop it from being blown to turn around in the opposite direction?)
A helicopter would offer steering power, and some thrust to counter what the wind is sending at you.

Overall, I think a small helicopter (or propelled aerial vehicle of some sort) would offer more stability.

There is a rule against damaging the terrain

[dexterpexter]

The only problem I could see with this is that driving through things was not seen as an acceptable solution by DARPA. It stipulated that the terrain and obstacles must be left unharmed. I think there are reasonable allowances made, such as running through "weeds" and leaving faint tire tracks.

Sending a bulldozer through something, however, would likely cause harm.

The motive behind this, if I get to guess, is that they are looking for a more covert vehicle. Something that has torn through the terrain and left chaos in its wake is more likely to be tracked/disabled than something that can quickly and nimbly navigate across the terrain.

I think that your idea is a fine idea, though. If they are looking at application for war situations and covert navigation is not an issue, I think that you are onto something.

When I first heard about the competition, that was my first reaction, too. Why not just create a tank and plow through the terrain along the most direct route? A review of the rules showed that they had already taken into consideration this solution and created a rule against it. I can see their reasoning, though.

I'm not so sure

[Einer2]

From what I know of the race course, these vehicles have to average 30 mph going cross-country through the desert. If it's anything like the terrain around the Tucson area, I'm not sure that I could average that without piling straight into a saguaro.

Re:I think they will

[jbrocklin]

Actually, I've read most of the technical papers that the teams were required to submit, and many of them did use a "track and assign danger levels" as a way of finding a best path (most used this as a way of keeping the vehicle inside the boundaries of the course - assign the off-course sections with infinite danger and the vehicle will never go there).

Overall, the majority of the problems that people were with unplanned problems, such as going up a hill and not switching down gears, stopping to check terrain, and then not being able to start back up again. Or getting a wheel stuck in a small ditch and not being able to get out (no friends to jump out of the car and push ya know).

I think that someone will win next year - or at least make it far enough that the logic part of the system will be proven effective. Sure there will be some little things that will hit just about every team, but I hope someone does a good enough job preparing that it will take a lot of little things to bring them down.

Mod me down

[wonkavader]

But while I'm interested in the beginning of that paragraph, and the end, I'm not gonna read what's in the middle.

The return key is your friend.

Re:couple of thoughts...

[Lije Baley]

Well, in response to your first concern, while it is certainly a shame that international teams can't contribute to solving this problem from their unique perspectives, it's probably not in the U.S.' security interest to encourage such projects abroad. Today's good friend can always become tomorrow's bitter enemy - this is not paranoia, it's history. If nothing else, it's simply good (internal) politics to try to keep the potential economic and scientific benefits within the U.S. For the record, I truly believe in diversity. Not just because it's fair or in vogue, but because I believe the best ideas come from the broadest set of perspectives. I think this has been key to the U.S.' technological success in the past, feeding off our "melting pot". It seems that the E.U. can rival this, and to compete with Asia, serious EU-US cooperation may be required to maintain any semblance of our present standards of living. Sorry, I digress...

As for your second concern(s), the basic argument for defense applies to the "help the military" question. As long as humans are human, there will be those who seek to take advantage of others, consciously or not, and we as individuals or groups can ignore that only at our own peril. History almost certainly reveals only a tiny fraction of those people(s) who were overrun and exterminated or assimilated by those who were aggressive and capable. Some in that 95% mentioned above may go on to use their good ideas to develop something which could threaten the U.S. and others, as a U.S. citizen I would expect us to research and develop any potentially unique military capability, so that we could at least understand and counter it.

Your fear of the Terminator/Matrix scenario is, IMHO, really premature at this point. Even if they succeed in the DARPA challenge, it will be a long time until we have roving, fully-autonomous-fire vehicles that have the capability to operate and maintain themselves, let alone their supporting infrastructure. The military has enough problems with friendly-fire as it is, some already due to poor machine decisions. I don't think they are so hot on turning ANYTHING (man or machine) loose with a weapon. I would guess the military's real direction with this stuff is to use autonomous machines to reduce communication link requirements and tedium for remote operation by humans - i.e. one man could be controlling vehicle A, targeting and firing weapons, while also being responsible for vehicles B and C, autonomously enroute to different recon sites, and vehicle C is currently out of contact due to terrain, weather, enemy jamming, equipment failure, etc, but returning or maneuvering to get back in touch.